The Exploratory Advanced Research Program

A Primer for Agent-Based Simulation and Modeling in Transportation Applications

Introduction

Agent-based modeling and simulation (ABMS) has been widely applied across a spectrum of disciplines by both researchers and practitioners. Examples of these disciplines include ecology, biology, business, economic science, computer simulation, social sciences, political science, policy, and military studies. Knowledge and applications of ABMS continue to expand and accumulate through rapid and in-depth research and development.

ABMS has been applied to a broad range of domains in transportation. These applications primarily fall into two methodological paradigms: individual-based models that study personal transportation-related activities and behavior, and computational (or system) methods that study a collaborative and reactive transportation system that exhibits intelligence by modeling a collection of autonomous decisionmaking of subsystem entities called agents. The former is closely related to the models for travelers' activities. The latter is typically scoped as a computational method in a distributed artificial intelligence (DAI) system, or a complex adaptive system (CAS), which is a powerful technique for simulating dynamic complex systems to observe emergent behavior. In research literature, it is common to see transportation studies crossing the boundary of the two categories but scoped with the same (or similar) term, agent-based, thus leading to conceptual confusion.

The goal of this primer is to review the historical aspects and the ongoing developments of ABMS in the interdisciplinary transportation areas, summarizing and clarifying the scope and key characteristics of past agent-based studies and to shed light on future potential research. Another scientific focus of this primer is to document a research effort that attempts to establish the relationship between the classical and ABMS-based route choice models. This effort aims to answer a scientific inquiry: Because both classical econometric models and ABMS are plausible in modeling individuals' route choice decision behaviors, there supposedly exist certain conditions and contexts at which both modeling paradigms exhibit comparable results. This inquiry, as an important step toward a better understanding of the classical econometric method and ABMS-based approaches, sheds light on the path forward for the development of a holistic modeling framework.

The objectives of this primer are to:

• Explain the basic concept of ABMS and various ABMS methodologies scoped in the literature.
• Review ABMS applications that have emerged in transportation studies in the past few decades.
• Describe the general ABMS modeling frameworks and commonly shared procedures exhibited in a variety of transportation applications.
• Outline the strength and weakness of ABMS in various transportation applications.
• Demonstrate, through a traveler's route choice decisionmaking process example, that ABMS exhibits certain modeling outcomes compared with classical approaches.

In reviewing ABMS applications in transportation, this primer serves to present and summarize the ABMS approaches that have been practiced in the transportation paradigm in the past few decades and depict the concept of ABMS as discussed in the literature. The applications described in this primer represent major recognized transportation systems and platforms that have been leveraging ABMS, rather than comprehensively reviewing products.

This primer is organized by chapters and sections. In chapter 1, fundamental concepts of ABMS are introduced, and both benefits and challenges of this methodology are presented. In chapter 2, the authors discuss how to model learning, and interactions in general, through an agent-based method derived from the individual-based human behavioral perspectives in the social science paradigm. In chapter 3, the authors briefly introduce several agent-based simulation software toolkits. These toolkits have the standardized modules, processes, libraries, and programming language that could be used conveniently to develop an agent-based model. In chapter 4, the authors discuss agent-based behavioral models that have been studied by the transportation community. Most of those models are individual-based models, in which agents are individual travelers. Those models have strong roots in activity-based transportation models. In chapter 5, the authors review agent-based system modeling in transportation problems where agents are intelligent, distributed, and autonomous subsystems. These subsystems (agents) interact with each other and model complex holistic performance and emergent behavior of the overall transportation system. In chapter 6, the authors seek to demonstrate that ABMS could be viable to model the individual traveler's route choice decisionmaking process, when compared with classical methods. In chapter 6, a behavioral model leveraging the social science methodology, called belief-desire-intention (BDI), is established in a bottom-up framework. This model attempts to formulate the mechanism of a traveler's complex route choice behavioral process as a collaborative and reactive result of the traveler's mindset and the network environment.

The target audiences of this primer are researchers and practitioners in the interdisciplinary field of transportation, who specialize or have an interest in social science models, behavioral models, activity-based transportation models, lane-use models, human factors, and artificial intelligence (AI) models with applications in transportation.